View Full Version : advice on capping daily show to xvid
sander815
7th May 2006, 08:23
i am capping a daily show running between 10-15 min. The source is pretty HQ mpeg2 streams from my satreceiver. That gets reduced to 400/300 Xvid avi, and should be as small as possible
I tried various filters in avisynth, but the result is still blocked and not very good with a datarate of around 600 kbps
Cana nyone advice me a decent script, that still gives me good speed?
see here (http://84.245.14.152/~movies/Nick.jpg) for an example of the raw material
sander815
7th May 2006, 18:26
this is the same frame, compressed:
http://84.245.14.152/~movies/Nick2.jpg
i am currently usding this script:
MPEG2Source("D:\movie.d2v")
ColorMatrix()
Crop(4,8,-4,-8)
TomsMoComp(1,5,1)
DeGrainMedian()
UnDot()
FluxSmoothST(5,8)
LanczosResize(416,312)
#BicubicResize(416,312,0,0.5)
aSharp(1,1)
ColorYUV(levels="pc->tv")
#saturation_v = 1.2
any suggestions on improving?
foxyshadis
7th May 2006, 21:38
Fluxsmooth is probably what's killing all those details. Everything else looks okay. (If it's a clean capture, ratchet down the degrainmedian and fluxsmooth strength though.) A source clip would help ascertain how much filtering needs to be done.
sander815
7th May 2006, 21:52
1 sene:
http://rapidshare.de/files/19885020/Nick_Talpa-05_05_2006-15_34-Talpa_vivienne_-_05_05_06_.ts.html
2nd scene
http://rapidshare.de/files/19885052/Nick_Talpa-04_05_2006-15_34-Talpa_vivienne_-_04_05_06_.ts.html
so, shot with HQ cameras(studio), and (i think) some lower quality cameras(school scene)
foxyshadis
8th May 2006, 00:22
I don't think you need fluxsmooth at all. It looks quite good with no filtering, and a bit better with undot°rainmedian, but fluxsmooth smears everything. asharp, oddly, also seems to remove some detail. It looks pretty good with "limitedsharpenfaster(dest_x=416,dest_y=312)" though.
sander815
8th May 2006, 15:42
what do i need to get limitedsharpen to work?
i thought i had it working, but VD crashes when the job is almost finished. Whhen i disable limitedsharpen everything is ok
Chainmax
8th May 2006, 19:56
If it's a clean source, use FFT3DFilter instead like this:
FFT3DFilter(sigma=x,bw=32,bh=32,bt=3,ow=16,oh=16)
with x=3 it seems to be about on par with DeGrainMedian(), so start with that and lower or increase it according to your needs.
If you want small size with great results I'd recommend you to try x264, but if you need Xvid (for standalone playback, for instance), try the eqm_v3ulr matrix and a bitrate of about 600-650kbps with a resolution of 448x336 (so you keep MOD16 dimensions). Is the source telecined or interlaced?
foxyshadis
8th May 2006, 20:17
http://avisynth.org/mediawiki/wiki/LimitedSharpen
You might need to update your masktools, and get the latest LimitedSharpenFaster.
sander815
8th May 2006, 21:47
If it's a clean source, use FFT3DFilter instead like this:
FFT3DFilter(sigma=x,bw=32,bh=32,bt=3,ow=16,oh=16)
with x=3 it seems to be about on par with DeGrainMedian(), so start with that and lower or increase it according to your needs.
If you want small size with great results I'd recommend you to try x264, but if you need Xvid (for standalone playback, for instance), try the eqm_v3ulr matrix and a bitrate of about 600-650kbps with a resolution of 448x336 (so you keep MOD16 dimensions). Is the source telecined or interlaced?
eqm_v3ulr matrix??
telecined or interlaced...no idea
Chainmax
8th May 2006, 23:14
eqm_v3_ulr is a custom matrix which you can find here (http://www.webalice.it/f.corriga/CQMs/eqm_v3ulr_rev3.xcm).
I suggest you to read Decomb's (http://www.neuron2.net/decomb/decombnew.html) readmes and tutorials, the IVTC tutorial (http://www.doom9.org/ivtc-tut.htm) and the capture guide (http://forum.doom9.org/showthread.php?t=78797).Those are a great source of starter knowledge and will explain telecining and interlacing with detail.
TomsMoComp(1,5,1) <- if source is fully interlaced
LanczosResize(320,240) <- if you want low resolution ..
then set xvid up to encode 1 pass quantizer 3.
Enable b frames with max consecutive 2.
sander815
9th May 2006, 12:42
i uploaded a small part with eqm_v3_ulr and this:
http://rapidshare.de/files/20010603/test.avi.html
enocoded in 2 passes
-
MPEG2Source("D:\test.d2v")
ColorMatrix()
Crop(4,8,-4,-8)
TomsMoComp(1,5,1)
DeGrainMedian()
UnDot()
LanczosResize(400,320)
#BicubicResize(416,312,0,0.5)
WarpSharp()
ColorYUV(levels="pc->tv")
please tell me if this can be improved, i can't say i am satisfied with this quality
Chainmax
9th May 2006, 16:30
A sample encode does not do any good at all because we can't assess the state of the source material. A small (10-20sec) unprocessed sample capture would be much better. A few tips though:
- Heed foxyshadis's advice and try LimitedSharpen.
- Read the guides I linked to, the source might be telecined and you'll be encoding redundant frames if you just deinterlace (and tomsmocomp is a very outdated deinterlacer anyway).
- Ditch Undot, using it along with DeGrainMedian is like putting an ant on top of an elephant.
sander815
9th May 2006, 18:35
i provided 2 samples in this thread, 1 inside studio and 1 outside
Chainmax
9th May 2006, 19:52
Didn't notice them, my bad.
vBulletin® v3.8.11, Copyright ©2000-2026, vBulletin Solutions Inc.